Research And Implementation Of Low Frequency Communication Signals Detection Technology

This paper mainly studies the recognition technology of conventional low-frequency modulation signal,focusing on the identification theory of single unknown signal,such as signal existence detection,carrier frequency estimation,modulation mode recognition,and makes the corresponding hardware implementation of the structure of DSP + FPGA.The specific contents are as follows:Firstly,the development,research status and basic theory of Blind Source Separation and Modulation Recognition technology are introduced.The principle,classification,uncertainty and evaluation index of Blind Source Separation(BSS)are briefly described.Combining instantaneous information,power spectrum and higher order spectrum,the general mathematical expressions and characteristics of six kinds of modulation signals in target signal set {AM,2ASK,BPSK,QPSK,2FSK,MSK} are presented.Secondly,the problem of hybrid signal separation with fewer observers than the source signal,i.e.blind source separation in undetermined scenarios,is studied.Three mature hybrid matrix estimation algorithms are introduced: K-means clustering algorithm,FCM clustering algorithm and angle retrieval clustering algorithm.The SSDP algorithm based on the mixed matrix to recover the known source signal is introduced,which can directly reflect the result of signal separation.The simulation results show that the generalized crosstalk errors and angle errors estimated by the angle retrieval clustering algorithm are less than one quarter of those estimated by K-means algorithm and FCM algorithm.Thirdly,based on the separated single unknown signal,the modulation recognition technology which is easy to implement by physical hardware is studied.The recognition process includes four steps: existence detection,carrier frequency estimation,feature parameter extraction,classifier classification and recognition.The flatness index of power spectrum is used to detect the existence of signals.When the signal-to-noise ratio is greater than-8dB,the detection success rate reaches 100%.The square spectrum estimation method estimates the carrier frequency,and the relative error is less than 0.1% under the signal-to-noise ratio of 0-30 dB.When the signal-to-noise ratio is greater than 5dB,the recognition rate of the whole system reaches more than 92%.Finally,the hardware based on the structure of FPGA + DSP is used to realize the algorithm of signal existence detection,carrier frequency estimation and modulation mode recognition.Real-time detection is carried out using the signal generated by the actual signal source.The results of recognition by feedback from PC show that the accuracy of existence detection is 100%,the relative error of carrier frequency estimation is less than 0.12% and the recognition accuracy of modulation mode is more than 93% for noise and six kinds of conventional low frequency modulation signals.The hardware implementation of the three functional modules is in line with expectations.